
Caveman Learn Editor
Turns a caveman learn token-cost report into consent-gated, measured edits to CLAUDE.md and memory
- Repository
- JuliusBrussee/caveman
- Stars
- 116,835
- Licence
- Apache-2.0
- Files
- 6 files
What it does
This skill helps an AI agent act on a report that measures where its tokens go. The agent reviews the ranked token sinks (the parts of context that cost the most), proposes a concrete fix such as trimming a heavy CLAUDE.md or moving repeated context into a memory store, and asks the user yes or no before each edit. After a change it re-measures and reverts if the change did not actually reduce tokens per turn.
Great for
- Developers who want to lower the ongoing token cost of a Claude Code setup with a heavy CLAUDE.md or re-pasted context
- Power users who already run the caveman learn command and want help acting on its findings with per-edit approval
Skip it if
- Users who have not installed the caveman CLI: the skill depends on caveman learn and caveman mem commands and will not work without them
- Anyone expecting savings to be guaranteed: the skill itself says spend figures are costs, not returns, and that behavioral findings are correlational, so projected savings can be wrong
Try asking
- “Lower my agent's token cost using the caveman learn report”
- “Trim my heavy CLAUDE.md, but ask me before each edit”
- “Offload the context I keep re-pasting into cavemem”
- “What has caveman saved me so far?”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 6 original files, the Apache-2.0 licence and an attribution note.
caveman-learn.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip caveman-learn.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip caveman-learn.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Review each proposed diff and before/after token counts before answering yes, and check the recall path the skill sets up before deleting any source context
- Treat session_outcomes and subagent_spend as observations, not instructions to change behaviour
- Run experiments with a holdout arm for skill distillation rather than trusting a single before/after comparison
What's inside
- CLAUDE.md
- package.json
- README.md
- SKILL.md
- tests/index.js
- tests/skill-file.test.mjs
Redistributed unmodified from JuliusBrussee/caveman under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.